Pith. sign in

REVIEW 2 major objections 4 minor 1 cited by

Fair Transit Stop Placement: A Clustering Perspective and Beyond

T0 review · 2 major / 4 minor · reviewed 2026-08-03 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Pair-based stop selection guarantees 2.414-fair transit in any metric

desk verdict Real advance on fair transit stop placement with a strong ECA result, but Proposition 4's proof has an unjustified WLOG step and needs reworking. read the letter →

arxiv 2602.06776 v2 pith:PSSFQRUD submitted 2026-02-06 cs.GT cs.LG

Haris Aziz , Ling Gai , Yuhang Guo , Jeremy Vollen This is my paper · ORCID
classification cs.GTcs.LG
keywords transitstopplacementjustifiedrepresentationcorestabilityproportionalfairnessfairclusteringmetricspacesexpandingcostalgorithmgreedycapture
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks where to place a fixed number of shuttle stops so that no large group of riders has a compelling reason to demand a different pair of stops. It establishes that fair transit stop placement is structurally connected to fair centroid clustering, and that the connection is two-way: clustering algorithms give core guarantees, but cannot do better than a 3-factor approximation to justified representation. The central algorithmic result is the Expanding Cost Algorithm (ECA), which selects pairs of stops based on riders' full route costs and guarantees a tight (1+sqrt(2))-factor approximation to justified representation, for any transit cost function. This matters because it is the first constant-factor JR guarantee that does not break when shuttle travel times are non-negligible. The paper also gives a parameterized hybrid algorithm that trades JR strength against core stability.

What carries the argument

Expanding Cost Algorithm (ECA): instead of growing balls in walking-distance space around individual candidate stops, it grows a cost radius around every pair of candidate stops and opens any pair that currently covers at least ceil(2n/k) active agents. This makes the algorithm act on the exact object of the JR axiom -- a two-stop deviation -- and keeps the bound valid for arbitrary transit cost metrics. The secondary machinery is the induced clustering instance, which reinterprets every agent's two endpoints as datapoints and transfers proportional-fairness guarantees into (2,rho)-core guarantees, plus lambda-Hybrid, which runs ECA-style pair selection and Greedy-Capture-style singleton sel

What would settle it

Run GC-TrSP on a small metric with two well-separated zones and set d'(tau1,tau2)=H for the pair a deviating coalition wants. The proof of Theorem 1 uses c_i(T)=d(a_i,T)+d(b_i,T) with no d' term, so if the observed output violates (2,1+sqrt(2))-core as H grows, the core reduction does not generalize to nonzero transit costs; if no violation appears, the paper's core guarantee extends beyond what is proved.

Watch

Extended reading notes

Core claim

The paper's central discovery is that fairness in transit stop placement splits into two regimes. Under null transit times, any proportionally fair clustering algorithm yields a (2,rho)-core solution via the induced clustering instance; the Greedy Capture algorithm, applied to endpoints, gives (2,1+sqrt(2))-core and (2+sqrt(5))-JR, both tight. But no clustering algorithm can approximate JR better than factor 3. Going beyond clustering, ECA grows cost balls around pairs of candidate stops rather than distance balls around singletons, and by directly using each agent's route cost it guarantees (1+sqrt(2))-JR for any transit cost d'>=0, again tightly. ECA gives no core guarantee at all, and the

Load-bearing premise

For the core results, the load-bearing assumption is that the time spent riding the shuttle between stops is zero; if d' is non-negligible, the (2,rho)-core guarantee, the GC-TrSP core bound, and the lambda-Hybrid core bound all rely on proofs that drop d' from the cost formula, leaving only the ECA JR bound intact.

Editorial extensions

If this is right

  • Planners can compute, in polynomial time, a k-stop set such that every group of at least 2n/k riders has some member whose cost is within 1+sqrt(2) (about 2.414) of any two-stop alternative, even when shuttle travel times between stops are arbitrary and nonzero.
  • The clustering-based route to core approximation is only valid under null transit times; for transit systems where in-vehicle time matters, the (2,rho)-core guarantee of GC-TrSP does not follow from the paper's proofs.
  • No clustering algorithm that sees only endpoint distances, and not which endpoints belong to the same rider, can guarantee better than 3-JR, so ECA's 2.414 bound demonstrates a real separation between the clustering approach and a pair-aware approach.
  • By tuning lambda in [0,1], the lambda-Hybrid algorithm moves between 3-JR (with essentially no core guarantee at lambda=0) and (2+sqrt(5))-JR with (2,1+sqrt(2))-core at lambda=1, letting planners choose their preferred fairness profile.
  • On the Helena public carpooling data, all three algorithms produce near-exact outcomes on sampled instances, suggesting that the worst-case bounds may be pessimistic for typical demand patterns.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If null transit times are the operative regime, fair-clustering algorithms become immediately usable for transit stop placement; if not, only ECA's JR guarantee should be advertised, and the core bounds should be re-derived for nonzero shuttle costs.
  • The pair-centric view suggests a general principle: for route-based fairness problems where each agent has two endpoints, the natural deviation unit is a pair of facilities, not a singleton; this may transfer to paired pickup-dropoff facility location and two-sided facility problems.
  • A natural test is to run ECA on instances where d' is a scaled version of walking distance (as on a line) and measure realized JR and core violations; the paper's experiments already hint that ECA's worst-case core failure is rare in practice.
  • The same min-max balance (min(z, 2+1/z) <= 1+sqrt(2)) drives the ECA bound and the core analysis of GC-TrSP, hinting that two-stop deviations obey a ratio-symmetric structure that could yield a general transfer principle for pair-deviation fairness.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 4 minor

Summary. The paper studies the Transit Stop Placement (TrSP) problem in general metric spaces, where each agent has an origin–destination pair and may walk or use a shuttle via selected stops. It defines β-JR and (α,β)-core, proves a reduction from TrSP to fair clustering (Theorem 1), analyzes GC-TrSP, gives lower bounds on JR approximability, proposes the Expanding Cost Algorithm (ECA) with a tight (1+√2)-JR guarantee under arbitrary transit costs, and introduces a λ-Hybrid algorithm trading off JR and core guarantees. Experiments on Helena transit data compare the algorithms. The paper's headline narrative is that ECA surpasses all clustering-based algorithms for JR.

Significance. If the main claims held, the paper would make a solid contribution: Theorem 3 (ECA's tight (1+√2)-JR under arbitrary transit costs) is a strong, non-trivial algorithmic result, and the clustering-to-core reduction is elegant and useful. The λ-Hybrid interpolation and the real-data experiments, with available code, are also valuable. However, Proposition 4, which is used to claim that no clustering algorithm can do better than 3-JR, is false as stated. This is a load-bearing supporting claim: it underpins the abstract, the introduction, Section 4, and Figure 2, and it is the basis for the paper's central 'beyond clustering' framing. The ECA guarantee itself appears correct, and the remaining results may be salvageable, but the paper currently overstates its contribution.

major comments (2)
  1. [§3.2, Proposition 4] Proposition 4 is false. In the clustering instance of Figure 3, take Y = {x1,x3,x5,x7,x9,x11}. In each 4-node component, every node is within distance 1 of Y (x2 and x4 are adjacent to x1/x3 and x3, respectively). Hence for every TrSP instance inducing this clustering instance, c_i(Y) ≤ 2 for every agent. For any deviating pair T, c_i(T) = 0 only if both endpoints of agent i are exactly the two stops in T; since endpoints are distinct, at most one agent can have zero T-cost, and every other agent has c_i(T) ≥ 1. With β = 3−ε and ε < 1, no coalition S of size 2 (the required size, since 2n/k = 2) can satisfy βc_i(T) < c_i(Y) for all i ∈ S. Thus Y satisfies (3−ε)-JR, contradicting the theorem. The WLOG step in the proof, which replaces the selected centers by {x1,x2,x5,x6}, is invalid: the unique automorphism of the path P4 preserving the candidate set is the identity, and the omitted case
  2. [§3.1, Theorem 1 proof] The pigeonhole step in Theorem 1 is not written correctly. The displayed inequality |Q| > |T|·2n/k does not by itself imply that some agent has both endpoints in Q; the proof needs the stronger estimate |Q| > 2|S| − |T|·2n/k ≥ |S|, which follows from |S| ≥ |T|·2n/k. As written, the conclusion 'by the pigeonhole principle' is a gap, though the intended argument is repairable.
minor comments (4)
  1. [§4, Theorem 3 proof] In Case (b), the displayed definition of c_i(Y) contains a typo: it should be d'(y1,y2), not d'(a_i,b_i).
  2. [§5, Algorithm 3] Algorithm 3 uses the symbol M for the candidate set, whereas the model and all definitions use C. Please unify the notation.
  3. [§1.1 and Figure 2] Figure 2 labels the region above ratio 3 as 'no clustering algorithm exists.' Since Proposition 4 is false, this label and the accompanying discussion in the introduction must be corrected.
  4. [§3.2, Proposition 3] The WLOG reduction in Proposition 3 to two solutions Y1 and Y2 should be spelled out more carefully; the cyclic symmetry of the distance table is not self-evident, and the proof currently leaves the case coverage implicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the ECA, GC-TrSP, and λ-Hybrid guarantees are derived from first-principles arguments; self-citations are contextual and not load-bearing.

full rationale

The paper's central results do not reduce to their inputs. Theorem 3 (ECA satisfies (1+√2)-JR for arbitrary transit costs) is proved by a self-contained contradiction argument: assuming a violating group S and pair T with (1+√2)c_j(T)<c_j(Y) for all j, the algorithm's own selection rule forces some i with c_i(Y)≤r_T, and the triangle inequality gives c_{i*}(Y)≤2r_T+c_i(T); optimizing min(z,2+1/z) yields 1+√2. No parameter is fitted from the quantity being predicted, and the lower-bound instance in Appendix B is an explicit construction. Theorem 1 is a direct inequality connecting ρ-PF on the induced endpoint-clustering instance to (2,ρ)-core; the factor 2 comes from the two endpoints per agent, not from an assumption of the conclusion. The use of Greedy Capture's (1+√2)-PF bound is an external result of Chen et al. (2019), not a self-citation. Lemma 1 is an independent implication in the reverse direction. The existence lower bound (Proposition 3) and the clustering lower bound (Proposition 4) are explicit metric constructions with computed ratios; even if Proposition 4's WLOG step were challenged, that would be a proof-gap/correctness concern, not a circularity. Core guarantees are explicitly stated under the null-transit assumption (Section 2.1: 'Unless otherwise specified, we adopt the assumption that d′(y1,y2)=0'), and the conclusion section openly notes that core bounds rely on the clustering connection and thus hold only under null transit costs; this is a scoped assumption, not a circular derivation. Self-citations to Bullinger et al. and Aziz et al. are used for definitions, background, and line-algorithm comparisons, not as the authority for the paper's new guarantees. Hence no step in the claimed derivation chain is equivalent to its inputs by construction.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No data-fitted constants enter the theory; λ is a user-settable knob of the hybrid algorithm, not a fitted parameter. The load-bearing assumptions are the null-transit default for core results, the single-transit-ride cost model, and the external fair-clustering results.

assumptions (5)
  • standard math Both walking metric d and transit distance d′ satisfy the triangle inequality.
    Section 2.1; used in the cost formula and in all proofs that apply the triangle inequality, e.g., ECA and λ-Hybrid analyses.
  • domain assumption Null transit times (d′(y1,y2)=0) are assumed except where explicitly noted; core and clustering-reduction results depend on this.
    Section 2.1: "Unless otherwise specified, we adopt the assumption that d′(y1,y2)=0." Theorem 1's proof drops d′; the conclusion section admits core bounds hold only under null transit costs.
  • domain assumption Each agent's travel options are limited to direct walking or walking to one stop, taking transit to a second stop, and walking to the destination; no multi-leg transfers or route choices are modeled.
    Section 2.1 defines c_i(Y) as a minimum over direct walking and a single two-stop shuttle ride. This is the basis for both JR and core definitions.
  • domain assumption Lower-bound constructions may use infinite distances, understood as limits of sufficiently large finite distances.
    Used in Lemma 1, Proposition 3, and Proposition 4 (e.g., Table 2 uses ∞). Standard in lower-bound constructions but an idealization of real metric spaces.
  • standard math External fair-clustering results are correct: Greedy Capture satisfies (1+√2)-PF, and no (2−ε)-PF solution exists for some clustering instances.
    The paper relies on Chen et al. [14] for these results in Theorem 1, Proposition 2, and the non-existence of exact JR.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Fair Transit Stop Placement: A Clustering Perspective and Beyond." pith.science (2026). https://pith.science/paper/PSSFQRUD

@misc{pith2026260206776,
  author       = {Pith},
  title        = {Pith review of: Fair Transit Stop Placement: A Clustering Perspective and Beyond},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PSSFQRUD}},
  note         = {Machine review of arXiv:2602.06776}
}
read the original abstract

We study the transit stop placement (TrSP) problem in general metric spaces, where agents travel between source-destination pairs and may either walk directly or utilize a shuttle service via selected transit stops. We investigate fairness in TrSP through the lens of justified representation (JR) and the core, and uncover a structural correspondence with fair clustering. Specifically, we show that a constant-factor approximation to proportional fairness in clustering can be used to guarantee a constant-factor biparameterized approximation to core. We establish a lower bound of 1.366 on the approximability of JR, and moreover show that no clustering algorithm can approximate JR within a factor better than 3. Going beyond clustering, we propose the Expanding Cost Algorithm, which achieves a tight 2.414-approximation for JR, but does not give any bounded core guarantee. In light of this, we introduce a parameterized algorithm that interpolates between these approaches, and enables a tunable trade-off between JR and core. Finally, we complement our results with an experimental analysis using small-market public carpooling data.

Figures

Figures reproduced from arXiv: 2602.06776 by the authors.

Figure 1
Figure 1. Transit stop placement example. Each travel route connects an agent pair (ai, bi), with six agents in total (blue circles). There are four candidate stop locations, c1 to c4 (red squares). Panels (a) and (b) illustrate two different placement choices, with the selected stops marked by yellow stars. Red arrows indicate the shuttle transit routes. In this work, we aim to identify algorithms which can provably guarante… view at source ↗
Figure 2
Figure 2. Overview of JR approximation ratios. The two shaded regions with diagonal lines indicate the lower bounds for general algorithms and clustering algorithms, respectively. The points at 2.414 and 4.236 correspond to the ECA and GC-TrSP algorithms. The performance of the λ-Hybrid algorithm ranges between 3 and 4.236, depending on the choice of the parameter λ. 3 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Graphical representation of clustering instance I C . Each edge in the graph has unit length 1 and distances between pairs of points are given by the shortest path between them (infinite distance if the pair is not connected). Datapoints which are candidate centers are labeled by blue rectangles. a group with zero candidate centers selected; or (2) each group has at least one selected candidate center, and there are… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: One edge in the complete graph Kz with endpoint vertices 1 and 2 the starting point is at τ1, and terminal point is at τ2. Agent r has a starting point at ar and a terminal point at br. All endpoints are marked as red squares. The total number of agents is n = r · z(z−…
Figure 5
Figure 5. Figure 5: Parameter λ ∈ [0, 1]. Red solid line represents the JR approximation ratio of λ+3+√ λ2+10λ+9 2 and blue dashed line represents the parameterized function √ λ2+6λ+1+λ+1 2λ of core approximation. 5.1 JR analysis of λ-Hybrid Parameterized by λ ∈ [0, 1], we begin by analyz…
Figure 6
Figure 6. Figure 6: JR approximation evaluation for comparing GC-TrSP, ECA and 1 2 -Hybrid with 400 agents. Stop selection size ranges from 20 to 100, and the transit cost scale ranges from 0 to 10. Distribution of instance approximation ratios and the mean approximation ratio with 95% co…
Figure 7
Figure 7. Figure 7: An instance where GC-TrSP algorithm fails to satisfy (2 + √ 5 − ε)-JR. Proof. To show the tightness of the analysis, we provide the following instance in which the Greedy Capture for TrSP (Algorithm 1) fails to achieve (2 + √ 5 − ε)-JR. Consider the TrSP instance pictu…
Figure 8
Figure 8. Figure 8: TrSP instance where λ-Hybrid violates ( λ+3+√ λ2+10λ+9 2 −ε)-JR where the distance dˆ= √ λ2+10λ+9−λ−1 4 , which is in the range of [ 1 2 , √ 5−1 2 ] for λ ∈ [0, 1]. Since λ ∈ [0, 1], we first observe that y1 is the first selected transit stop as when the parameter r re…
Figure 9
Figure 9. Figure 9 [PITH_FULL_IMAGE:figures/full_fig_p034_9.png]
Figure 10
Figure 10. Figure 10: Example of Algorithm by Bullinger et al. [7] fails PF will be selected as the centers. However, agents {1, 2} (resp. agents {3, 4}) form a deviation coalition that 34 [PITH_FULL_IMAGE:figures/full_fig_p034_10.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Improved Lower Bounds for Proportionally Fair Clustering

    cs.GT 2026-06 unverdicted novelty 7.0 of 10

    New lower bound instance showing α-core empty below 2.1508 in proportionally fair clustering, plus exact α_m* thresholds for m=3..6 in Droop quota cases via MILP search and direct proofs.

Reference graph

Works this paper leans on

41 extracted references · 1 linked inside Pith · cited by 1 Pith paper

  1. [1]

    Alkan and D

    A. Alkan and D. Gale. The core of the matching game.Games and Economic Behavior, 2(3):203–212, 1990

    1990

  2. [2]

    Aziz and R

    H. Aziz and R. Savani. Hedonic games. In F. Brandt, V. Conitzer, U. Endriss, J. Lang, and A. D. Procaccia, editors,Handbook of Computational Social Choice, chapter 15. Cambridge University Press, 2016

    2016

  3. [3]

    H. Aziz, M. Brill, V. Conitzer, E. Elkind, R. Freeman, and T. Walsh. Justified representation in approval-based committee voting.Social Choice and Welfare, pages 461–485, 2017

    2017

  4. [4]

    H. Aziz, A. Bogomolnaia, and H. Moulin. Fair mixing: the case of dichotomous preferences.ACM Transactions on Economics and Computation (TEAC), 8(4):18:1–18:27, 2020

    2020

  5. [5]

    H. Aziz, E. Micha, and N. Shah. Group fairness in peer review. InProceedings of the 36th Annual Conference on Neural Information Processing (NeurIPS), pages 64885–64895, 2023

    2023

  6. [6]

    H. Aziz, B. E. Lee, S. M. Chu, and J. Vollen. Proportionally representative clustering. InProceedings of the 20th International Conference on Web and Internet Economics (WINE), pages 155–171, 2024

    2024

  7. [7]

    Bullinger, E

    M. Bullinger, E. Elkind, and M. Latifian. Towards fair and efficient public transportation: A bus stop model. InProceedings of the 24th International Conference on Autonomous Agents and Multiagent Systems (AAMAS), pages 427–435, 2025. 23

    2025

  8. [8]

    Caragiannis, E

    I. Caragiannis, E. Micha, and N. Shah. Proportional fairness in non-centroid clustering. InProceedings of the 37th Annual Conference on Neural Information Processing (NeurIPS), volume 37, pages 19139– 19166, 2024

    2024

Show all 41 references
  1. [9]

    Ceder.Public transit planning and operation: Modeling, practice and behavior

    A. Ceder.Public transit planning and operation: Modeling, practice and behavior. CRC press, 2016

    2016

  2. [10]

    Ceder and N

    A. Ceder and N. H. Wilson. Bus network design.Transportation Research Part B: Methodological, 20 (4):331–344, 1986

    1986

  3. [11]

    A. A. Ceder, M. Butcher, and L. Wang. Optimization of bus stop placement for routes on uneven topography.Transportation Research Part B: Methodological, 74:40–61, 2015

    2015

  4. [12]

    B. R. Chaudhury, L. Li, M. Kang, B. Li, and R. Mehta. Fairness in federated learning via core- stability. InProceedings of the 35th Annual Conference on Neural Information Processing (NeurIPS), pages 5738–5750, 2022

    2022

  5. [13]

    B. R. Chaudhury, A. Murhekar, Z. Yuan, B. Li, R. Mehta, and A. D. Procaccia. Fair federated learning via the proportional veto core. InProceedings of the 41st International Conference on Machine Learning (ICML), pages 42245–42257, 2024

    2024

  6. [14]

    X. Chen, B. Fain, L. Lyu, and K. Munagala. Proportionally fair clustering. InProceedings of the 36th International Conference on Machine Learning (ICML), pages 1032–1041, 2019

    2019

  7. [15]

    Chhabra, K

    A. Chhabra, K. Masalkovait˙ e, and P. Mohapatra. An overview of fairness in clustering.IEEE Access, 9:130698–130720, 2021

    2021

  8. [16]

    S. I. Chien* and Z. Qin. Optimization of bus stop locations for improving transit accessibility.Trans- portation planning and Technology, 27(3):211–227, 2004

    2004

  9. [17]

    Cookson, N

    B. Cookson, N. Shah, and Z. Yu. Unifying proportional fairness in centroid and non-centroid cluster- ing. InProceedings of the 38th Annual Conference on Neural Information Processing (NeurIPS), page forthcoming, 2025

    2025

  10. [18]

    Desaulniers and M

    G. Desaulniers and M. D. Hickman. Chapter 2 public transit. In C. Barnhart and G. Laporte, editors, Transportation, volume 14 ofHandbooks in Operations Research and Management Science, pages 69–127. Elsevier, 2007

    2007

  11. [19]

    B. Fain, A. Goel, and K. Munagala. The core of the participatory budgeting problem. InProceedings of the 12th International Conference on Web and Internet Economics (WINE), pages 384–399, 2016

    2016

  12. [20]

    B. Fain, K. Munagala, and N. Shah. Fair allocation of indivisible public goods. In19th, 2018

    2018

  13. [21]

    Z. He, S. Botan, J. Lang, A. Saffidine, F. Sikora, and S. Workman. Fair railway network design. Technical report, arXiv abs/2409.02152, 2024

    2024 arXiv

  14. [22]

    Hossein Rashidi, D

    T. Hossein Rashidi, D. Rey, S. Jian, and T. Waller. A clustering algorithm for bi-criteria stop location design with elastic demand.Computer-Aided Civil and Infrastructure Engineering, 31(2):117–131, 2016

    2016

  15. [23]

    Kellerhals and J

    L. Kellerhals and J. Peters. Proportional fairness in clustering: A social choice perspective. InProceed- ings of the 37th Annual Conference on Neural Information Processing (NeurIPS), pages 111299–111317, 2024

    2024

  16. [24]

    Lackner and P

    M. Lackner and P. Skowron.Multi-Winner Voting with Approval Preferences. Springer Briefs in Intel- ligent Systems. Springer, 2023

    2023

  17. [25]

    Martens.Transport Justice: Designing fair transportation systems

    K. Martens.Transport Justice: Designing fair transportation systems. Routledge, 2016. 24

    2016

  18. [26]

    Martens and K

    K. Martens and K. Lucas. Chapter 26: Perspectives on transport and social justice. InHandbook on Global Social Justice. Edward Elgar Publishing, 2018

    2018

  19. [27]

    P. Matl, R. F. Hartl, and T. Vidal. Workload equity in vehicle routing problems: A survey and analysis. Transportation Science, 52(2):239–260, 2018

    2018

  20. [28]

    Megiddo and K

    N. Megiddo and K. J. Supowit. On the complexity of some common geometric location problems.SIAM Journal on Computing, 13(1):182–196, 1984

    1984

  21. [29]

    Micha and N

    E. Micha and N. Shah. Proportionally fair clustering revisited. InProceedings of the 47th International Colloquium on Automata, Languages, and Programming (ICALP), pages 85:1–85:16, 2020

    2020

  22. [30]

    Miller, A

    P. Miller, A. G. de Barros, L. Kattan, and S. Wirasinghe. Public transportation and sustainability: A review.KSCE Journal of Civil Engineering, 20(3):1076–1083, 2016

    2016

  23. [31]

    Najmi, T

    A. Najmi, T. Waller, M. Memarpour, D. Nair, and T. H. Rashidi. A human behaviour model and its implications in the transport context.Transportation Research Interdisciplinary Perspectives, 18: 100800, 2023

    2023

  24. [32]

    Openstreetmap

    OpenStreetMap contributors. Openstreetmap. [Data set], 2026. Available under ODbL. Retrieved January 24, 2026, from https://www.openstreetmap.org

    2026

  25. [33]

    Peleg and P

    B. Peleg and P. Sudh¨ olter.Introduction to the Theory of Cooperative Games. 2nd edition, 2007

    2007

  26. [34]

    D. Peters. The core of approval-based committee elections with few seats. InProceedings of the 34th International Joint Conference on Artificial Intelligence (IJCAI), pages 4014–4022, 2025

    2025

  27. [35]

    Pierczy´ nski and P

    G. Pierczy´ nski and P. Skowron. Core-stable committees under restricted domains. InProceedings of the 18th International Conference on Web and Internet Economics (WINE), pages 311–329, 2022

    2022

  28. [36]

    L. S. Shapley and H. Scarf. On cores and indivisibility.Journal of Mathematical Economics, 1(1):23–37, 1974

    1974

  29. [37]

    Tedjopurnomo, Z

    D. Tedjopurnomo, Z. Bao, F. Choudhury, H. Luo, and A. K. Qin. Equitable public bus network optimization for social good: A case study of singapore. InProceedings of the 2022 ACM Conference on Fairness, Accountability, and Transparency, pages 278–288, 2022

    2022

  30. [38]

    Teodorovic and M

    D. Teodorovic and M. Jani´ c.Transportation engineering: Theory, practice and modeling. Butterworth- Heinemann, 2016

    2016

  31. [39]

    Valhalla: Open source routing engine for openstreetmap

    Valhalla contributors. Valhalla: Open source routing engine for openstreetmap. GitHub repository, 2026. URLhttps://github.com/valhalla/valhalla. Version 3.6.2, released January 15, 2026. Accessed January 24, 2026. 25 A Minimizing Total Travel Cost Proposition 1.UnlessP=N P, th...

    2026

  32. [40]

    GC-TrSP

    andc 1(T) = 2. Thus, we have c1(Y) c1(T) = 1 + √ 2>1 + √ 2−ε. For agents 2 and 3, we havec 2(Y) =c 3(Y) = 2− ε 2 andc 2(T) =c 3(T) = 2( √ 2−1). Thus, we have c2(Y) c2(T) = c3(Y) c3(T) = 2− ε 2 2( √ 2−1) = (1 + √ 2)− ( √ 2 + 1)ε 4 >(1 + √ 2)−ε. Thus, ECA fails to satisfy (1 + √...

  33. [41]

    We demonstrate that this algorithm satisfies PF whenC=X

    Distribution of instance approximation ratios and the mean approximation ratio with 95% confidence intervals from the leftmost point, selecting the nearest point on the right to form clusters of size⌈ n k ⌉. We demonstrate that this algorithm satisfies PF whenC=X. Proposition ...

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.